The effect of Staphylococcus aureus on the persistence of Pseudomonas aeruginosa due to aggregation in cystic fibrosis airway infections

Summary

Pseudomonas aeruginosa (Pa) and Staphylococcus aureus (Sa) co-infection has been associated with worse clinical outcomes in people with cystic fibrosis (pwCF); however, the nature of their interaction is poorly understood. This was a 3-year prospective study of pwCF with chronic Pa infection. Pa, Sa,and staphylococcal protein A (SpA) were visualized by confocal microscopy in sputum samples obtained at baseline, exacerbation, after antibiotic treatment,and recovery. Forty-five participants were enrolled. At exacerbation, Pa biovolume and aggregate number increased in children, as well as geospatial distancing of Pa from Sa. With IV antibiotics, Pa (0.5–20 μm diameter) decreased significantly more than large Pa aggregates (>20 μm). Increasing sputum SpA concentrations were correlated with higher Pa biovolume in children (p = 0.04, r = 0.54). Within CF airways, Pa and Sa occupy spatially distinct niches but may interact through SpA production in children. Large Pa aggregates persisted despite antibiotics, indicating sustained infection.

Graphical abstract

Highlights

• •P. aeruginosa and S. aureus occupy spatially distinct niches
• •At exacerbation, P. aeruginosa biovolume and aggregate number increase
• •Large P. aeruginosa aggregates persist despite antibiotic treatment
• •Higher staphylococcal protein A levels are associated with greater P. aeruginosa biovolume

Immune response; Bacteriology; Microbiofilms

Introduction

It is well-documented that the behavioral dynamics of polymicrobial communities in human infections differ from single-species infections, with important consequences for disease severity and treatment response.¹ Chronic airway infection in cystic fibrosis (CF) exemplifies the following: CF airways are persistently colonized by multispecies bacterial communities that drive recurrent infection and inflammation.² The two most common pathogens co-isolated in such infections are Pseudomonas aeruginosa (Pa) and Staphylococcus aureus (Sa), co-occurring in 20%–50% of people with CF (pwCF) in cohorts worldwide.^{3,4,5,6,7,8,9,10} Depending on experimental conditions, CF clinical isolates of Pa and Sa may coexist or spatially segregate, cooperate, or antagonize one another, which reshape community composition, alter antibiotic susceptibility, and may influence clinical outcomes.¹¹

Despite overwhelming evidence of interspecies association, the net clinical impact of Pa-Sa co-infection on CF patient outcomes remains unclear.¹² Retrospective clinical studies in CF have been inconsistent: some report no significant difference between Pa mono-infections compared to Pa-Sa co-infection,^5,6 others report poorer outcomes associated with Pa mono-infection irrespective of Sa,¹³ and a significant number of studies show an association between Pa-Sa co-infection and worse clinical outcomes, characterized by increased lung inflammation, enhanced antimicrobial resistance, lower pulmonary function, and higher frequency of pulmonary exacerbation (PEx).^4,7,8,9,10

Modeling limitations likely contribute to the conflicting findings of Pa-Sa interaction. Assays that combine both species in a single in vitro system often fail to recapitulate the complex CF lung environment. These include heterogeneous nutrient availability, host immune responses, spatial structure of lung niches, prior antibiotic exposures, and variation in strain genotype or virulence-each a potential confounder of Pa-Sa associations in pwCF.^2,14 Because standard in vitro approaches rarely recreate clinically relevant community compositions, discrepancies between studies underscore both the difficulty of modeling polymicrobial infections and the challenge of translating single-species antimicrobial susceptibility data into treatment success for multispecies infections.¹⁵

Within the CF lung microbiome, bacterial communities may interact directly through physical or chemical means. We have previously shown in vitro that staphylococcal protein A (SpA), a virulence factor secreted by Sa, binds to the Pa exopolysaccharide, Psl, leading to aggregation and tobramycin resistance.¹⁶ We thus hypothesized that Sa interacts with Pa through SpA production, resulting in antimicrobial resistance and increased Pa, which may contribute to worse clinical outcomes in CF. The goal of our study was to visualize and measure Pa, Sa, and SpA directly in sputum samples from pwCF, using an adapted microbial identification after passive CLARITY Technique (MiPACT) in combination with fluorescent in situ hybridization (FISH) and immunofluorescence monoclonal antibody labeling, to clarify how Pa and Sa coexist or compete in CF airways.¹⁷ To do so, we conducted a 3-year prospective observational study at three CF centers of pwCF with chronic Pa infection using sputum samples obtained longitudinally at baseline (B), exacerbation (E), on antibiotic treatment (T) and at recovery (R).^18,19,20 Many in vitro studies have attempted to model clinically relevant Pa-Sa interactions; to our knowledge, this is the first large scale longitudinal study to use MiPACT with FISH to directly visualize bacterial communities in patient respiratory samples to answer this question.

Results

Characteristics of study subjects

Forty-five participants (12 pediatric and 33 adult) were enrolled and followed for a median of 2 years; 91 sputum samples were analyzed from the cohort. Demographic characteristics of the cohort are summarized in Table 1. Overall, the pediatric group contributed more sputum samples per subject (median 4.5 vs. 2.0) and had longer follow-up (median 14.5 vs. 5.7 months) than the adult group. The mean duration of chronic Pa infection was 2.3 years in children compared to 10.0 years in adults, with some adults exceeding 20 years. Sa co-infection was more frequent in the pediatric cohort than the adult cohort (75.0% vs. 54.5%).

Table 1.

Demographics of study cohort at enrollment

	Pediatric (N = 12)	Adult (N = 33)
Age (years), mean (SD)	12.5 (3.0)	35.8 (11.3)
Female, n (%)	8 (66.7)	18 (54.5)
DeltaF508 homozygous, n (%)	4 (33.3)	10 (30.3)
Pancreatic Insufficient, n (%)	10 (83.3)	31 (93.9)
Baseline ppFEV1, median (range)	101 (71–109)	75.4 (41–113.2)
Total number of exacerbations	16	22
Number of exacerbations/patient, mean (SD)	1.3 (1.6)	0.7 (1.2)
Number of sputum samples/patient, median (range)	4.5 (1–11)	3.0 (1–15)
Duration of chronic P. aeruginosa infection, n (%)
0–5 years	9.0 (75.0)	8.0 (24.2)
6–10 years	3.0 (25.0)	12.0 (36.4)
>10 years	0.0 (0.0)	13.0 (39.4)
S. aureus co-infection, n (%)	9 (75.0)	19 (57.6)
Concomitant Medications, n (%)
CFTR modulator therapy	6 (50.0)	18 (54.5)
Chronic inhaled antibiotics	10 (83.3)	24 (72.7)
Chronic azithromycin	3 (25.0)	15 (45.5)
Hypertonic saline	5 (41.7)	12 (36.4)
Dornase alfa	6 (50.0)	12 (36.4)

ppFEV₁, percent predicted forced expiratory volume in 1 s.

CFTR, cystic fibrosis transmembrane conductance regulator.

Baseline measures of P. aeruginosa visualized in sputum samples

At baseline, we compared Pa biovolume, number of aggregates, maximum aggregate size, and surface-to-biovolume ratio, as visualized by confocal laser scanning microscopy (CLSM) in sputum samples from adult and children participants; adults showed a greater number of aggregates in sputum than children (p = 0.02; Figure 1). There were no significant differences in these measures between participants receiving cystic fibrosis transmembrane conductance regulator (CFTR) modulator therapy compared to those not (Figure 2). Accordingly, subsequent analyses were stratified by age group only. There were no significant differences in Pa measures between Pa mono and Pa-Sa co-infected subjects, in either children or adults (Figure 3).

Figure 1.

Baseline, denoted by B, measures of Pa in sputum from pediatric (black circles) versus adult (white circles) patients

Each circle represents a sputum sample.

Grouped analyses of Pa biovolume (A), number of aggregates (B), maximum aggregate size (C), and surface-to-biovolume ratio (D).

Median values are marked with black horizontal lines. Comparison between pediatric and adults showed a significantly greater number of Pa aggregates in adults (p = 0.02) using Mann-Whitney non-parametric t test. ∗p < 0.05.

Figure 2.

B Pa measures in sputum from patients on CFTR modulator therapy (black circles) versus off modulators (white circles)

Grouped analyses of Pa biovolume (A), number of aggregates (B), maximum aggregate size (C), and surface-to-biovolume ratio (D).

Median values are indicated with black horizontal lines.

Comparison between patients on and off CFTR modulator therapy showed no significant differences using non-parametric t test.

Figure 3.

B measures of Pa mono-infection versus Pa-Sa co-infection in sputum from pediatric (black circles) and adult (white circles) patients

Grouped analyses of Pa biovolume (A), number of aggregates (B), maximum aggregate size (C), and surface-to-biovolume ratio (D).

Median values are marked with horizontal black lines.

Comparison between Pa mono- and Pa-Sa co-infection showed no significant differences using Mann-Whitney non-parametric t test.

Changes from baseline to onset of PEx

At the onset of exacerbation, we noted significant increases in Pa biovolume (p = 0.03) and the number of Pa aggregates (p = 0.01) in children; Pa measures in adults did not differ from B to E (Figures 4 and 5). Although the absolute abundance of Sa remained low compared to Pa and was largely stable between B and E, Sa maximum aggregate size (p = 0.03) and surface-to-biovolume ratio (p = 0.045) increased significantly between these two time points in the children (individual subject data shown in Figure S1). These changes were not observed using CFU/mL or PCR measures or in SpA sputum concentrations (Figure S2). To determine whether Pa mono versus Pa-Sa co-infection status affected the changes observed from B to E, we performed a two-way ANOVA within each group to assess for interaction (Figure 6); there was only a trend (p = 0.06) for interaction based on mono vs. co-infection status for the B to E changes in Pa biovolume in the adults.

Figure 4.

B versus exacerbation (day 0) (denoted by E) measures of Pa and Sa in sputum from pediatric (black circles) and adult (white circles) patients

Grouped analyses of Pa biovolume (A), the number of aggregates (B), maximum aggregate size (C), and surface-to-biovolume ratio (D).

From B to E, pediatric samples exhibited significant increases in Pa biovolume (p = 0.03) and the number of Pa aggregates (p = 0.01), while measures of Sa in the same sputum were relatively stable.

In pediatric sputum, Sa maximum aggregate size (p = 0.03) and surface-to-biovolume ratio (p = 0.045) also increased.

Adults showed no significant change in Pa or Sa measures between B to E. Each symbol represents an individual sample; median values are marked with horizontal black lines.

Statistical comparisons were performed using Mann-Whitney non-parametric t test. ∗p < 0.05.

Figure 5.

Representative CLSM images (20x) of sputum from a pediatric patient at B (A) and E (B) showing increase in Pa biovolume and number of aggregates

Pa (green) and Sa (blue) aggregates. SpA is shown in magenta.

The length of grided squares is 24 μm.

Figure 6.

B versus E (day 0) measures of Pa mono-infection versus Pa-Sa co-infection in sputum from pediatric (black circles) and adult (white circles) patients

Grouped analysis of Pa biovolume (A), number of aggregates (B), maximum aggregate size (C), and surface-to-biovolume ratio (D).

Each symbol represents an individual sample; median values are marked with horizontal black lines.

Comparison between Pa mono- and Pa-Sa co-infection groups was done by two-way ANOVA to assess for interaction based on co-infection status with Bonferroni post-tests.

There was a trend p = 0.06 for interaction based on mono vs. co-infection status for the B to E changes in Pa biovolume (A) in the adults. ∗p < 0.05.

To determine whether the proximity between Pa and Sa changed between B to E, we conducted a geo-spatial analysis of the two taxa in three pediatric and three adult sputum samples. At B, there was a negatively skewed distribution of distances between Pa and Sa, whereas at E, the distribution of Pa to Sa was broader and more evenly distributed across larger distances (Figure 7). At B, the mean distance of Pa to the nearest Sa was 43.8 μm (SD 35.8), whereas at E, it increased to 87.4 μm (SD 46.2). Figure S3 provides representative images demonstrating the measurement technique.

Figure 7.

Geospatial distance between the centroid of Pa aggregates to the nearest Sa aggregate in sputum collected at B and E (day 0)

Z score histogram at B (black) and E (gray), with mean values 43.8 μm (SD, 35.8 μm) and 87.4 μm (SD, 46.2 μm), respectively.

Dataset includes z stack images from pediatric (n = 3) and adult (n = 3) samples, in Pa-concentrated areas, where Pa and Sa were present in the same field of view (20× Objective).

Grouped comparison between B and E showed a negatively skewed distribution in the distance between Pa to the nearest Sa, where at B, Pa, and Sa were closer together; at E, Pa, and Sa distance measures were more evenly distributed compared to B.

Changes with antibiotic treatment of PEx

To evaluate response to antibiotic therapy, we analyzed Pa and Sa measures at the onset of exacerbation, 7–14 days into IV antibiotic treatment and at the follow up clinic visit (recovery). Significant decreases in Pa biovolume, aggregate number and maximum aggregate size were observed from E to T and reductions in biovolume and number of aggregates were maintained at R in both children and adults. Significant decreases in Sa measures were noted from E to R in the adults only (Figure 8) (individual subject data shown in Figure S4 and sputum CFU/mL, PCR and SpA concentration data shown in Figure S5). The responses to IV antibiotic therapy as measured from E to T were not affected by Sa co-infection status, as determined using a two-way ANOVA (Figure 9).

Figure 8.

E (day 0) versus IV treatment (T) and recovery (R) analysis of Pa and Sa biovolume (A), the number of aggregates (B), maximum aggregate size (C), and surface-to-biovolume ratio (D) in pediatric (black circles) and adult (white circles) sputum

From E to T, significant declines are shown in Pa biovolume, number of aggregates and the maximum aggregate size in children and adults.

From E to T, Sa measures showed no significant changes in children or adults. From E to R, a significant decline in all Sa measures is shown in adults; no trend shown in children.

Mann-Whitney t test was performed for statistical analysis. Each symbol represents an individual sample; median values are indicated with black lines. ∗p < 0.05, ∗∗p < 0.01 and ∗∗∗p < 0.001.

Figure 9.

E (day 0) versus IV T analysis of Pa mono-infection versus Pa-Sa co-infection in pediatric (black circles) and adult (white circles) sputum

Analysis of Pa biovolume (A), number of aggregates (B), maximum aggregate size (C), and surface-to-biovolume ratio (D).

Median values are indicated with horizontal black lines. Comparison between mono and co-infected groups done by two-way ANOVA to assess for interaction based on co-infection status with Bonferroni post-tests.

There was no statistically significant interaction based on infection status for the E to T changes for any of the Pa measures. ∗p < 0.05, ∗∗p < 0.01.

To determine whether IV antibiotic treatment was preferentially eliminating planktonic and smaller aggregates of Pa, we compared counts of Pa (0.5–20 μm in diameter) with counts >20 μm from E to T using a threshold of 20 μm. This threshold was chosen based on the mean of the Pa aggregate diameter in sputum and in the differences observed in the distribution of aggregate size before and after antibiotic therapy (Figure 10). Pa 0.5–20 μm decreased by 5-fold (p = 0.0005) whereas large Pa aggregates >20 μm decreased by only 2-fold (p = 0.012), with a statistically significant greater drop in Pa 0.5–20 μm compared to >20 μm (p = 0.002, Mann-Whitney test). Thus, a greater proportion of the sputum Pa population was composed of large aggregates post IV treatment (30%) compared to pre-treatment (17%). In contrast, Sa aggregate counts did not change significantly for populations 0.5–10 μm or >10 μm (based on average aggregate diameter in sputum samples) from E to T (Figure 10).

Figure 10.

Changes in the size of Pa and Sa aggregates in sputum at PEx (day 0) and treatment (day 7)

Grouped comparisons between a) Pa of small size (0.5–20 μm) and large size aggregates (>20 μm) and Sa of small size (0.5–10 μm) and large size aggregates (>10 μm).

Paired analysis of the dataset includes sputum from pediatric (black circles) and adult (white circles) patients. Median values are indicated with horizontal black lines.

From E to T, the proportion of small Pa aggregates is shown to significantly decline by 5-folds (p = 0.0005); large Pa aggregates decline by 2-folds (p = 0.012).

No significant changes in the proportion of small and large Sa aggregates shown from E to T. (B) Pie chart representing the proportion of large and small aggregates at E and T for Pa and Sa.

Representative CLSM images of adult patient, PS2005 showing (C) small and large Pa (green) aggregates at E and (D) the elimination of small Pa aggregates while large Pa aggregates remain at T.

Very few Sa aggregates (blue) shown in the Pa concentrated region. SpA is shown in magenta. The length of grided squares is 24 μm. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

Of note, significant changes in Pa and Sa measures were not observed with oral antibiotic therapy (Figures S6 and S7).

In vitro effects of SpA on Pa

Although we did not observe co-localization of the SpA antibody and Pa by CLSM, as we hypothesized, we did note a significant relationship between increasing sputum SpA concentrations measured by ELISA and higher Pa biovolume detected by FISH in sputum samples from children (p = 0.04, r = 0.54) (Figure 11); there was no correlation between Pa biovolume and Sa biovolume, CFU/mL or PCR (Figure S8).

Figure 11.

Pa biovolume visualized in sputum by FISH graphed versus SpA measured in the same sputum by ELISA

A positive correlation shown in pediatric (black circles); no relationship shown in adults (white circles).

Pearson correlation performed for statistical analysis.

To further examine the effects of SpA on Pa, we selected three Pa isolates from the children sputum samples with high Pa biovolume (by CLSM) and high sputum SpA concentrations (by ELISA), in contrast with three samples with low Pa biovolume and low sputum SpA concentrations. These isolates were grown as biofilms using a chamber cover glass model for 24 h then exposed to varying concentrations of SpA and tobramycin for 24 h before calculating CFU/mL. Figure 12C shows that the addition of as little as 50 pg/mL of SpA resulted in high-level tobramycin resistance (1,000 μg/mL). When each Pa isolate was exposed to its respective SpA concentration measured in the sputum sample, a relationship was observed between increasing SpA concentrations and higher Pa CFU/mL (Figure 12D).

Figure 12.

Pa isolates (n = 6) were grown as biofilms in a glass slide chamber wells for 24 h then treated for an additional 24 h with varying concentration of SpA and (A) LB alone, (B) LB + 500 μg/mL tobramycin (T500), and (C) LB + 1,000 μg/mL tobramycin (T1000)

Pa isolates were selected from subjects with low sputum Pa biovolume measured by FISH and low sputum SpA concentrations measured by ELISA (light gray bars) as well as from subjects with high sputum Pa biovolume and high sputum SpA concentrations (dark gray bars); PAO1 (black bars).

(D) the same Pa isolates were grown as biofilms in the chamber wells for 24 h then exposed to the specific SpA concentration corresponding to the SpA concentration measured by ELISA in the sputum from which the Pa isolate was recovered, as well as varying concentrations of tobramycin for an additional 24 h before CFU counts were measured.

A positive correlation between Pa CFU/mL and SpA concentrations was noted in LB alone (gray circles) (p < 0.01, r = 0.93), LB + T500 μg/mL (black circles) (p < 0.05, r = 0.81) and LB + T1,000 μg/mL (open circles) mL (p = 0.08, r = 0.76). Statistical analyses performed by two-way ANOVA (A, B, and C) and Pearson correlation (D). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

Discussion

In the present study, we identified a relationship between increasing sputum SpA concentrations and greater Pa biovolume visualized in pediatric CF sputum (an effect recapitulated in vitro), and at the time of exacerbation, significant increases in Pa biovolume and aggregate counts. In both cohorts, there were greater spatial distancing of Pa from Sa at exacerbation, and IV antibiotic therapy preferentially eliminated planktonic and smaller Pa aggregates but were less effective at clearing larger aggregates. This study was novel as it used MiPACT with FISH to directly visualize polymicrobial interactions in patient sputum samples.

We hypothesized that increasing SpA concentrations would correlate with increased Pa in CF airways, based on our prior in vitro finding that SpA binds to Pa Psl leading to aggregation and tobramycin resistance.¹⁶ Although we did observe this positive association in pediatric CF sputum, the association was relatively weak (r = 0.54) and there was considerable heterogeneity of the sample. In addition, this relationship is correlative and does not establish a direct causal effect in vivo. Notably, we did not note this correlation in adults, likely due to their lower overall incidence of Sa infection. The difference in the duration of chronic Pa infection between children and adults likely resulted in isolate characteristics that may also have contributed to results varying based on age. Additionally, we could not demonstrate significant co-localization of SpA antibody with FISH labeled Pa visualized by CLSM. Because SpA is both cell-associated and secreted by Sa and binds the Fc region of IgG to inhibit opsonophagocytosis, endogenous IgG or immune complexes in CF sputum may be capping Pa bound SpA, thereby preventing anti-SpA antibody binding, as previously shown in vitro by Armbruster et al.²¹ An alternate explanation for the lack of detectable co-localization may be the variability in the timing of SpA antibody binding between samples. It is possible that the binding observed in our fixed sputum samples, collected at specific timepoints, did not capture this variability. We found that SpA may be present in CF sputum even in the absence of Sa culture positivity, which may account for the discrepancy with the Sa co-infection effect on Pa biovolume and aggregate measures visualized by CLSM in our study. In vitro, we also noted the relationship between increasing SpA concentrations and higher Pa measured by CFU/mL using selected children clinical isolates. Minimal amounts of SpA (as low as 50 pg/mL) were required to achieve maximal tobramycin resistance, without a dose dependent effect. The association between SpA and Pa can partially be explained by isolate-level differences. Specifically, Pa isolates recovered from sputum samples with higher SpA concentrations were able to grow better in the presence of tobramycin, even in the absence of SpA; conversely, Pa isolates recovered from sputum samples with low SpA did not grow well in presence of tobramycin, suggesting an intrinsic isolate effect (Figure 12).

We also observed rare co-localization of Sa and Pa in sputum samples from both children and adults with CF and chronic Pa infection. The average distance between Pa and Sa was approximately 44 μm during baseline visits. As previously shown by Rudkjobing et al., Pa and Sa aggregates exist in distinct microbial communities when visualized in CF sputum samples.²² It is important to note that in this current study, chronic Pa infection was an inclusion criterion which meant that Pa was present in the airways for at least a year, allowing Pa and Sa to develop adaptations permitting co-existence.^11,12,23,24 Limoli et al., for example, demonstrated that production of alginate by Pa, a common phenotype of chronic infection, promotes co-existence with Sa in an in vitro model of CF respiratory infection.²⁴ Thus, these findings may not be generalizable to acute infection. Importantly, Sa and Pa may interact differently depending on the clinical stage of the host.²⁵ At the time of exacerbation in our study, the distance between these two organisms increased to approximately 87 μm. This, together with the absence of direct Pa-Sa contact in our samples, suggests that diffusible factors rather than contact-dependent interactions may contribute to the spatial changes. Barraza et al. demonstrated a similar distancing between Pa and Sa in vitro upon elimination of Pa’s ability to produce the anti-staphylococcal small molecule HQNO (2-heptyl-4-hydroxyquinolone N-oxide).²⁶

During PExs, this study demonstrated significant increases in Pa biovolume and number of aggregates as well as increases in maximal Sa aggregation size from baseline in children with CF as visualized by CLSM; these changes were not detected by CFU count or PCR. Identifying a microbial signature at the onset of PEx has been challenging in CF. Previous studies using molecular methods such as PCR have not been able to show significant changes in either overall bacterial or Pa loads across exacerbation stages, even when sputum samples were treated with propidium monoazide to remove dead bacterial cells prior to DNA extraction.^27,28,29 Of note, these prior studies were all done in adults with CF whereas we demonstrated significant microbial changes only in CF children, who are known to have greater microbial diversity, less dominated by canonical CF pathogens, compared to adults.^30,31 Visualization of bacteria directly in situ using MiPACT-FISH has the advantage of maintaining bacterial structures such as aggregates which is not possible when using homogenized sputum samples for 16 S rRNA sequencing, PCR or colony counts. During PExs, sputum typically becomes more voluminous and purulent, but all the microbial measures detected by FISH in this study were reported per 1 mm sputum slice across all samples, suggesting that the increased number of Pa aggregates noted may be playing a role in the onset of exacerbations.

In addition, IV antibiotic therapy was associated with significant decreases in Pa biovolume and number of aggregates, as well as maximal aggregate size in both adult and children with CF and these changes were also seen using CFU/mL and PCR. Prior clinical trials have demonstrated significant drops in sputum Pa density as measured by CFU/mL following IV antibiotics.³² Reassuringly, the drops in Pa biovolume and number of aggregates found in this current study were maintained at the one month follow up visit (R); similar responses were not observed with oral antibiotic courses. Despite the decrease in overall Pa, however, large aggregates (>20 μm) were less effectively cleared compared to smaller Pa aggregates, resulting in an increased proportion of larger Pa aggregates with IV antibiotic treatment. Several investigators have demonstrated that Pa in aggregates resist killing by antimicrobials such as tobramycin,^33,34,35,36 as well as impair neutrophil phagocytosis.³⁷ The exopolysaccharide matrix of aggregates slows the penetration of antimicrobials.³⁸ Bacteria within aggregates also have slower growth rates leading to the ability to survive otherwise lethal concentrations of antibiotics, known as tolerance.³⁹ The charge of components within the matrix can also impede certain charged antibiotics (e.g., tobramycin) and can lead to interactions with host factors.^40,41 To our knowledge, our study is the first to directly demonstrate in patient samples, longitudinal enrichment of >20 μm Pa aggregates following IV antibiotic therapy, suggesting that Pa eradication may only be possible by disrupting bacterial aggregates to revert susceptibility to antibiotics and neutrophil phagocytosis.⁴¹

In conclusion, Pa and Sa coexist in spatially distinct niches in CF airways but may interact through the production of SpA, which is associated with increased Pa burden in children. Pediatric PExs are marked by greater Pa biovolume and aggregate number and PExs overall are linked with increased geospatial distancing between the microbes. Although IV antibiotics decrease overall Pa burden, ineffective clearance of large aggregates likely contributes to the persistence of Pa in pwCF. Important clinical observations can thus be made using the novel MiPACT technique combined with FISH to directly image bacterial communities within patient samples, circumventing the limitations of in vitro modeling.

Limitations of the study

There were several limitations to this study. Visualization of Sa in sputum using FISH was challenging given the difficulty in penetrating the bacterial cell wall with the probes, requiring careful adjustment of lysostaphin and lysozyme concentrations depending on the purulence of individual sputum samples. These technical issues may have limited the sensitivity of Sa visualization by FISH; however, we also assessed Sa using CFU/mL and PCR. In addition, although the study enrolled both CF adults and children with chronic Pa infection from three CF centers, there was a relatively small sample size (45 subjects), limiting the power of inter group comparisons between mono- and co-infected adults and children. Similarly, there were few PExs overall, given that half the study patient population was on CFTR modulator therapy. Nevertheless, significant changes were noted from baseline to exacerbation timepoints with respect to bacterial biovolume and aggregate measures. The introduction of CFTR modulator therapy has also decreased the incidence of Pa-Sa co-infection, limiting the generalizability of our study findings.⁴² Furthermore, the lack of advanced lung function measures, such as multiple breath washout and magnetic resonance imaging, limited the assessment of the clinical impact of these findings. Although comprehensive strain-level characterization through sequencing would provide valuable insight into clonal persistence, such analyses were beyond the scope and funding of this present study. Lastly, although microscopy can be inherently subjective, our systematic protocol of examining several views, from multiple slices by two independent observers, for each sputum specimen minimized potential bias.

Resource availability

Lead contact

Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Valerie Waters (valerie.waters@sickkids.ca).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• •The data that support the findings of this study are available from the lead contact upon request.
• •No custom code was generated in this study.
• •Any additional information required to reanalyze the data reported in this study is available from the lead contact upon request.

Acknowledgments

This study was funded by CF Canada (1017927) and the Cystic Fibrosis Foundation (6600100518). A.J.M. was supported by a research fellowship from CF Canada.

Author contributions

A.J.M., S.E., G.F., S.A., and L.F. performed the experiments. A.S., E.T., V.J.W., and D.N enrolled the patients and obtained sputum samples and clinical data. A.J.M., S.E., Y.C.W.Y., W.H.D., I.M., E.T., D.N., and V.J.W. contributed to the study design, data interpretation and analysis, and writing of the manuscript.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-Protein A antibody, Mouse Monoclonal (clone SPA-27, purified from hybridoma cell culture)	Sigma-Aldrich	Cat# SAB4200745
Biological samples
Sputum samples	SickKids Hospital (Toronto, ON), St. Michael’s Hospital (Toronto, ON), and the McGill University Health Center (Montréal, QC).	N/A
Clinical bacterial strains recovered from sputum samples	SickKids Hospital (Toronto, ON), St. Michael’s Hospital (Toronto, ON), and the McGill University Health Center (Montréal, QC).	N/A
Critical commercial assays
Protein A ELISA Kit	Abcam	Cat# ab215542
DNeasy PowerSoil Pro Kit	Qiagen	Cat# 47016
Software and algorithms
Comstat (software 2.1)	Technical University of Denmark	https://www.comstat.dk/
Imaris (software 10.1.1)	Oxford Instruments	https://imaris.oxinst.com/versions/10-1
GraphPad PRISM 10 (San Diego)	Dotmatics	https://www.graphpad.com/
Volocity (software 6.3)	PerkinElmer	https://www.volocity4d.com
Other
PseaerA Probe	Eurofins Genomics	[AminoC6+Alexa 488]GGTAACCGTCCCCCTTGC)
Staaur Probe	Eurofins Genomics	[AminoC6+Alexa 4647]GAAGCAAGCTTCTCGTCCG

Experimental model and study participant details

This was a three-year (2022-2025) prospective, longitudinal observational study at three CF clinics: SickKids Hospital (Toronto, ON), St. Michael’s Hospital (Toronto, ON), and the McGill University Health Centre (Montréal, QC).

The inclusion criterion was pwCF presenting with chronic Pa infection, defined as more than 50% of respiratory cultures Pa-positive in the preceding 12 months or on chronic suppressive inhaled antibiotics.^18,19 Sputum and clinical data were collected through follow-up, including during PEx, defined as an increase in respiratory symptoms requiring treatment with intravenous (IV) or oral antibiotics.^43,44 Poor sputum quality was rejected based on the following criteria: liquid consistency, defined by the absence of visible plugs⁴⁵; presence of hemoptysis (interferes with MiPACT); and specimen weight <0.1g. In total, sputum samples from forty-five participants (12 pediatric and 33 adult) were collected and analyzed from the cohort. Demographic characteristics of the cohort are summarized in Table 1.

Sputum samples were classified using an adapted BETR framework based on medical record review²⁰: B, baseline (routine clinic visit); E, PEx (day 0); T, treatment with IV antibiotics (days 7–14); and R, recovery within 60 days post IV therapy. For PEx managed with oral antibiotics, recovery samples (r) were obtained within 60 days post-treatment as these patients typically return to clinic only after completing their antibiotic course. All antibiotic courses are detailed in Table S1. Sputum specimens were further categorized as Pa mono-infection or Pa-Sa co-infection based on culture positivity at B and E.

Ethics approval

The study was approved by the SickKids Research Ethics Board (REB #1000079038), the St Michael’s Hospital Research Ethics Board (REB #22-172) and the McGill University Health Center Research Ethics Board (REB #017-1626) in accordance with the Declaration of Helsinki.

Method details

Sputum processing and visualization of bacteria in sputum

Expectorated sputum was collected in a sterile container and immediately divided into two equal portions. One portion was fixed in 4% paraformaldehyde (PFA) for 24 h then processed using an adapted MiPACT, as previously described.^17,46,47 In brief, fixed sputum was washed with 1× phosphate-buffered saline (PBS) and sectioned into small slices (∼5 mm in diameter) to be embedded into hydrogels. Individual sections were placed into separate wells of an eight-chamber cover glass slide (Nunc Lab-Tek; Thermo Fisher Scientific, Mississauga, ON), each containing 250 μL hydrogel mixture. The cover glass slide was then placed in a BD GasPak EZ container (VWR, Mississauga, ON) with an anaerobic pack and incubated at 37 °C for 3 h to allow polymerization. Following polymerization, sputum–hydrogels were cleared in 8% sodium dodecyl sulfate (SDS) for 14-30 days at 37 °C. After clearing, samples were washed with 1× PBS and stored at 4 °C in 0.01% (wt/vol) sodium azide until confocal laser scanning microscopy (CLSM). The second half of sputum was mixed with 10% sputolysin (wt/vol) in 1X PBS and vortexed until homogeneous. The homogenate was divided into four equal aliquots for the following analyses: (1) quantification of Pa and Sa by serial dilution and plating onto blood agar to calculate colony-forming unit (CFU)/mL⁴⁸ (species identification were confirmed by MALDI-TOF); (2) quantification of Pa⁴⁹ and Sa (Thermo Fischer) by quantitative real-time PCR (RT-PCR) and (3) quantification of SpA by enzyme-linked immunosorbent assay (ELISA) as per manufacturer’s specifications (Abcam). Although we tried to perform all tests on every sputum sample, depending on the amount of sample, this was not always possible and therefore not every sample has corresponding values for all tests.

Fluorescent in situ hybridization and confocal laser scanning microscopy

In the portion of sputum processed by MiPACT, ex vivo visualization of Pa and Sa was performed by three-dimensional (3D) confocal laser scanning microscopy (CLSM) following FISH with species-specific probes: PseaerA (Pa; green) and Staaur (Sa; blue) (Eurofins Genomics).⁵⁰ SpA was visualized using a monoclonal antibody against SpA (magenta) (Sigma-Aldrich). For each sputum sample, a minimum of 6 (1mm in diameter) slices were taken and imaged by 2 independent observers. Normalizing analyses to a standardized 1 mm sputum section was specifically intended to control for variability in sputum volume and purulence across samples, including those collected during exacerbation. The mean of 6 slices for each bacterial measure was calculated for each sample. Briefly, each slice was incubated in MilliQ water with 0% to 75% lysozyme concentration (depending on the purulence of the sputum sample) and 0.6% lysostaphin concentration at 37 °C for 30 mins. Then, the slices were washed with 1X PBS and were incubated with 500 μL of hybridization buffer containing 4 μL of 100 μM PsearA (green) and Stauur probes (blue) (Eurofins) at 45 °C. After rinsing in wash buffer, slices were incubated in 1 mL of fresh wash buffer for 6 hours at 47 °C. Post-hybridization, slices were washed with 1× PBS and incubated in 2% bovine serum albumin (BSA) solution overnight (18–24 hours) at room temperature in the dark. The following day, slices were incubated with 50 μg/mL anti-SpA antibody conjugated to Alexa594 (red) (Mouse Monoclonal clone SPA-27, Sigma-Aldrich) in 2% BSA/PBS solution for 6 hours at room temperature. The slices were washed again with 1× PBS and incubated in 250 μL of refractive index matching solution (RIMS) at room temperature with gentle rotation. All images were acquired using Quorum confocal laser scanning microscopy (CLSM), using a 20X objective on a Zeiss AxioVert 200M Microscope. Image z-stacks were acquired then analyzed in Volocity (software 6.3) to quantify biovolume, total number and maximum aggregate size for each species. An aggregate was defined as greater than 1 μm in diameter (minimum size of 2 Pa bacteria). Aggregate segmentation was performed using voxel connectivity based object identification following intensity thresholding of the confocal z-stacks. Thresholding was applied individually to each image to ensure accurate detection of visible aggregates while avoiding signal oversaturation. Specifically, the offset threshold for each image was adjusted to include all discernible aggregates without incorporating background noise. For size based quantification, a minimum object volume threshold of 0.5 μm³ was first applied to determine the total number of aggregates greater than this size. A second threshold of 20 μm³ was then applied to quantify aggregates exceeding this larger volume. The number of aggregates between 0.5 μm³ and 20 μm³ was calculated by subtracting the latter value from the former one. In addition, the Comstat 2.1 plugin to ImageJ software was used to quantify aggregation level (measure by surface-to-biovolume ratio).⁵¹ Spatial proximity between the species was assessed in Imaris (software 10.1.1) by calculating the 'spots to spots closest distance' from the centroid of a Pa aggregate to the centroid of the nearest Sa aggregate. In this computation, Imaris also generates randomized spots (stimulating the same number as observed) to establish expected values, which allows users to see whether the data falls outside the randomized envelope (shaded region). Imaris refers to this computation as a “randomized null hypothesis. A total of 95 aggregates at E were analyzed: E sample #1 = 25 aggregates; E sample #2= 42 aggregates; E sample #3= 28 aggregates. A total of 114 aggregates at B were analyzed: B sample #1: 42 aggregates; B sample #2: 48 aggregates; B sample #3: 24 aggregates.

FISH probes specificity for S. aureus/ P. aeruginosa

To determine the specificity of the Stauur FISH probe binding to Staphylococcus aureus and its ability to discriminate against coagulase-negative Staphylococci (CoNS) isolated from CF patients, we conducted in vitro testing using planktonic cultures; n=12 CoNS isolates were grouped into three cocktails and cultured in LB medium. The first cocktail contained S. epidermidis (337/42), S. lugdunensis (364/33), S. epidermidis (340/53), and S. epidermidis (342/136). The second cocktail included S. epidermidis (331/70.3), S. capitis (327/97.2), S. capitis (332/85), and S. hominis (339/97). The third cocktail comprised S. hominis (334/018), S. haemolyticus (268/15), S. haemolyticus (337/05), and S. saprophyticus (342/27). Following planktonic growth, 50 μL (concentration 10⁶ CFU/ml) of each isolate was spotted onto glass slides and fixed with 4% paraformaldehyde (PFA). Fluorescent in situ hybridization (FISH) was then performed as previously described,⁴⁶ with minor modifications. Briefly, each slide was incubated with 500 μL of hybridization buffer containing 4 μL of 100 μM Stauur probe (blue) (Eurofins) at 45 °C. After rinsing in wash buffer, slides were incubated in 1 mL of fresh wash buffer for 6 hours at 47 °C. Post-hybridization, slides were washed with 1× PBS and incubated in 2% bovine serum albumin (BSA) solution overnight (18–24 hours) at room temperature in the dark. The following day, slides were washed again with 1× PBS and incubated in 250 μL of refractive index matching solution (RIMS) at room temperature with gentle rotation. All images were acquired using Quorum confocal laser scanning microscopy (CLSM), using a 20X objective on a Zeiss AxioVert 200M Microscope. The images in the z-stack were captured at a distance of 0.3 μm in order to obtain the biofilm depth. Volocity software was used for acquisition and analyses of all the images. The Stauur probe did not cross-react with any coagulase negative Staphylococci species (Figure S9). Additionally, n=43 sputum hydrogels with undetectable Sa (0 copies/mL by RT-PCR) were visualized, and no Sa signal was observed, confirming the specificity of the probe. To determine the sensitivity of the Stauur probe, n=1 clinical isolate of Sa was similarly cultured planktonically, serial dilutions performed and imaged using CLSM as above (Figure S10). The limit of detection (where several cells could be visualized reliably) for the Staaur Probe was determined to be approximately 10⁴ cells/mL. A similar protocol was used to assess the detection limit of the PsearA probe (Eurofins) for Pseudomonas aeruginosa. As previously reported,⁴⁶ the PsearA probe demonstrated a detection limit of 10⁴ cells/ml; the specificity is previously reported.⁴⁷

Anti-Spa antibody specificity

To determine the specificity of the anti-Protein A antibody (Mouse Monoclonal, IgG1 isotype, Sigma Aldrich) for S. aureus -secreted protein A (SpA), and to confirm its lack of cross-reactivity with secretions from coagulase-negative staphylococci (CoNS) isolated from CF patients, we performed in vitro biofilm experiments. Briefly, biofilms of S. aureus (clinical isolate) and n=6 CoNS strains (S. epidermidis (337/42), S. lugdunensis (364/33), S. capitis (327/97.2), S. hominis (339/97), S. haemolyticus (268/15), and S. saprophyticus (342/27)) were grown in Nunc Lab-Tek II glass slide chambered wells (Thermo Fisher Scientific, Mississauga, ON). Each isolate was cultured overnight in 4 mL of lysogeny broth (LB) at 37 °C with shaking at 225 rpm. A 40 μL aliquot of the overnight culture was diluted into 4 mL of fresh LB (BioShop) and grown to early log phase (OD₆₀₀ = 0.1). Subsequently, 220 μL of each culture was added to individual wells of an 8-chambered cover-glass slide and incubated for 24 hours at 37 °C to allow biofilm formation. After incubation, the medium was removed and replaced with 200 μL of either LB alone or LB supplemented with 50 μg/mL anti-SpA antibody conjugated to Alexa594 (red) (Mouse Monoclonal clone SPA-27, Sigma-Aldrich). The slides were incubated for an additional 24 hours at 37 °C. The following day, the medium was gently removed and replaced with 200 μL of Syto-9™ live-cell fluorescent stain in LB, followed by a 45-minute incubation at room temperature. Wells were then washed with LB, and 200 μL of fresh LB was added prior to imaging. All images were acquired using Quorum confocal laser scanning microscopy (CLSM), using a 20X objective on a Zeiss AxioVert 200M Microscope. The images in the z-stack were captured at a distance of 0.3 μm in order to obtain the biofilm depth. Volocity software was used for acquisition and analyses of all the images.

SpA was successfully visualized in the biofilm of the Sa clinical isolate, as evidenced by magenta fluorescence (Figure S11). In contrast, all CoNS strains showed no detectable signal in the magenta channel, indicating the absence of SpA and confirming the specificity of the anti-SpA antibody. Additionally, a negative control sputum hydrogel (0 pg/mL SpA measured by ELISA) was also visualized. No detectable SpA signal was observed, confirming the specificity of the antibody (Figure S12).

In vitro measurement of Pa biofilm growth in presence of SpA

P. aeruginosa isolates (PAO1 and n=6 clinical isolates) were selected based on their biofilm characteristics: either high P. aeruginosa biovolume with high SpA concentration (PS1001-C3, PS1004-E1(2^nd) and PS1007-C2) or low P. aeruginosa biovolume with low SpA concentration (PS1006-C1, PS1008-C1 and PS1009-C5). These isolates were grown as biofilms for 24 h in the wells of an eight-chamber cover glass slide (Fisher Scientific) and then exposed for an additional 24 h to varying concentrations of SpA and tobramycin, as previously described.⁵² Briefly, each P. aeruginosa isolate was grown overnight in 4 mL of lysogeny broth (LB) with shaking (225 rpm) at 37 °C. 40 μL of the overnight culture was diluted in 4 mL of fresh LB medium (BioShop) and grown to an early log phase of 0.1 optical density (OD) at 600 nm. Subsequently, 220 μL of this culture was added to each well of an 8-chambered cover-glass slide and incubated for 24 h at 37 °C for biofilm growth. The following day, the media was removed from each well and replaced with 200 μL of either LB alone or LB supplemented with varying concentrations of SpA [50 pg/mL, 100 pg/mL, 500 pg/mL 1000 pg/mL as well as the isolate-specific SpA concentration determined via ELISA in sputum] and/or tobramycin [500 μg/mL or 1000 μg/mL], depending on the experimental condition. The chambered cover-glass slide was then incubated for another 24 h at 37 °C. The following day, media was gently removed and was briefly washed with 200 μL of 1xPBS. The adhered biofilm was then scraped using a pipette tip for about one min and resuspended in 200 μL of 1xPBS. The resulting suspension was serially diluted and plated on blood agar plates. Plates were incubated for 24 h at 37 °C and the number of P. aeruginosa colonies were recorded as CFU/ml. All experiments included 3 biological replicates, each with 3 technical replicates and the means were calculated.

Quantification and statistical analysis

Continuous variables were compared between groups using non-parametric t-tests; a 2-way ANOVA was used to assess for interaction based on infection status (mono- versus co-infection) when assessing changes from B to E and E to T. Significance was defined as p<0.05; analyses done on GraphPad PRISM 10 (San Diego).